Difference in p-value significance between Weighted and Unweighted Unifrac Distances

[AndrwOnar]

Hello community,


I'm getting confused about how to interpret my results for beta-diversity.
Indeed comparing two groups I get a significant p-value for Bray Curtis and  Unweighted UniFrac but not for the Weighted.
Thus I suppose that I have different OTUs (bray curtis) that do not share a similar phylogenetic background (Unweighted) ...and??
how can I interpret the fact that the Weighted is not significant? Thay are different but in similar percentage? I wonder it does not make sense?
May you help me in find a correct interpretation? Does anyone have experienced the same results condition?

Thanks

[Embriette]

Hi Andrw,

Unweighted and weighted UniFrac are providing information on two different things. Unweighted takes into account presence/absence, while weighted takes into account abundance also. This certainly can yield different p-values.

Bray Curtis is a non-phylogenetic metric that like weighted UniFrac takes into account abundance. However, because Bray Curtis is a non-phylogenetic metric and weighted UniFrac is a phylogenetic metric that takes into account branch length on the phylogenetic tree, very small branch lengths on the tree might explain why something that is significant with Bray Curtis is insignificant with weighted UniFrac.

Thanks!

Embriette

[AndrwOnar]

Hi Embrietta,

Thank you for your explication, I'm not very sure that this could explain my situation because as you said both Unweighted and Weighted are phylogenetic and I get a significant p-value for the Unweighted while a not significant result for the Weighted so  I don't really understand how that could be possible.
It's like to say that my two groups are phylogenetically distant (unweighted significant) but thay became closer if we considere relative abundace...
what do you think?

I have  a personal theory about that situation: probably the most abundant sequences have a very strong influence in the cluster analysis....I mean if i consider just the Unweighted thay are distant under a phylogenetically point of view, but when i consider the relative abundance probably the abundance of the shared taxon is "stronger" than that of the non-shared taxon thus givin a no-significant result for Weighted Unifrac...but that's just a personal theory